Building a 24/7 Patient Intake System for Under €500/Month

A €45,000/year enterprise CRM deployment I was called in to audit had a 14-hour average TFCR. The clinic was paying more for software than for the coordinator it had replaced, and getting worse results than a €300/month stack would have delivered. The price of a patient intake system has nothing to do with its performance.

Last Updated: 20260407T0

I’ve built intake systems for clinics in Istanbul across hair transplant, dental, and cosmetic surgery. The question I get most often from clinic operators is whether a properly automated intake system requires a significant technology investment. The answer is no. The stack I deploy for mid-tier Turkish clinics costs between €280 and €480 per month depending on lead volume and AI usage, and it outperforms every enterprise alternative I have seen in production in this market.

What follows is the architecture, component-by-component cost breakdown, and deployment logic for a 24/7 intake system that a real clinic can run without a dedicated technology team.

What Is the Architecture and How Do the Components Connect?

The system has four functional layers, each handled by a different component in the stack. Understanding how they connect is more important than understanding any individual tool.

Layer one is the message reception layer. Evolution API connects to a WhatsApp Business number and exposes a webhook that fires every time a new message arrives. This webhook triggers an n8n workflow. The clinic’s WhatsApp number, whether a new number or an existing one migrated to the API, is the front door. Every patient message, at any hour, enters the system through this layer within seconds of being sent.

Layer two is the AI pre-qualification layer. The n8n workflow receives the message payload, extracts the patient’s message text, and sends it to OpenAI’s API with a structured prompt that has been configured for the clinic’s specific procedure mix. For a hair transplant clinic, this prompt identifies DHI versus Sapphire FUE interest, graft estimate requests, source country, and budget signals. For a dental clinic, it identifies veneer versus implant versus orthodontic interest. The AI returns a structured JSON response that includes the detected procedure type, the patient’s apparent language, and a summary of their inquiry.

Layer three is the response and CRM layer. n8n takes the AI’s structured output and does two things simultaneously. It sends a personalized first-response WhatsApp message to the patient, written in their detected language, referencing their specific procedure interest, and including a soft qualifying question. It also writes a new patient record to Supabase and creates a conversation in Chatwoot, pre-populated with every data point collected so far. By the time a coordinator opens Chatwoot, they see a labeled, categorized conversation with patient origin, procedure interest, and inquiry summary already visible.

Layer four is the human handoff layer. Chatwoot shows coordinators an inbox ordered by lead temperature and response urgency. Hot leads, patients who have confirmed a procedure interest and asked about pricing or availability, are surfaced first. Coordinators do not wade through unqualified inquiries. They engage with conversations that the system has already moved to a qualified state.

How Does the System Handle Multiple Languages and Time Zones?

1. Language Detection and Multilingual Response

The largest source markets for Turkish medical tourism, the UK, France, Germany, the Gulf states, each arrive in a different language. A coordinator who speaks English and Turkish cannot manage Arabic, French, and German inquiries without lag. The AI layer handles this without additional cost. The OpenAI prompt includes a language detection instruction, and the response template library in n8n holds first-response messages pre-written in English, French, German, Arabic, and Turkish. When a French patient sends an inquiry at 2am Istanbul time, they receive a French-language response within 90 seconds. No coordinator is involved.

2. Overnight and Weekend Coverage

In my experience with Istanbul clinics, the highest-value leads often arrive outside European business hours, because patients in the UK and Germany are researching medical procedures in the evening, after work. A clinic that closes at 6pm Istanbul time is effectively invisible to this segment. The automated intake system has no off-hours state. The n8n workflows run continuously on the VPS infrastructure. Weekend and overnight leads receive the same response quality as business-hours leads, and they enter the Chatwoot queue for coordinator follow-up when the team comes online the following morning, already pre-qualified, already in the CRM, not waiting for a first response.

3. Volume Scaling Without Cost Scaling

A manual intake operation scales linearly with volume, more leads require more coordinators. The automated system does not. OpenAI API costs are per-token, not per-seat. At 300 leads per month with an average of 500 tokens per pre-qualification interaction, the AI cost is approximately €45 to €60 per month regardless of whether those leads arrive simultaneously or spread across the month. The n8n VPS handles several hundred concurrent workflow executions without performance degradation. Chatwoot self-hosted handles thousands of conversations without additional licensing cost. A clinic growing from 100 to 300 leads per month does not need to increase its infrastructure spend materially.

What Does Deployment Actually Look Like for a Clinic?

The deployment process takes three to four weeks for a clinic starting from zero automation. Week one covers infrastructure setup: provisioning the VPS servers, installing n8n and Chatwoot, connecting Evolution API to the clinic’s WhatsApp number, and creating the Supabase schema. Week two covers workflow build: the intake trigger, the AI pre-qualification logic, the response templates in all relevant languages, and the CRM write operations. Week three covers testing with real lead traffic, starting with a small volume and validating that responses are accurate, CRM records are populating correctly, and Chatwoot handoff is working as expected. Week four covers coordinator training and go-live.

The clinic does not need a technical team to maintain the system after deployment. n8n workflows run autonomously. Chatwoot is a managed interface that coordinators use like any other messaging tool. The only ongoing technical task is monitoring for Evolution API instance health: WhatsApp connections occasionally require reconnection, which takes two minutes.

Revenue Leakage before this system is typically 40 to 60 percent of the lead pipeline. After full deployment, it drops to under 15 percent. For a clinic processing 150 leads per month at an average closed procedure value of €1,500 and a pre-system conversion rate of 10 percent, moving to 25 percent conversion represents 22 to 23 additional closed patients per month, or approximately €33,000 in incremental monthly revenue, against a system cost of under €400/month.

What Is the Underlying Principle Behind the Sub-€500 Intake Stack?

The enterprise software market for medical clinics is built on the assumption that complexity and cost are correlated with capability. In medical tourism intake, they are not. The capability a Turkish clinic needs is specific: respond instantly, qualify systematically, populate the CRM without human input, and hand off to coordinators with full context. None of these requirements need a €5,000/month CRM platform or a €50,000 custom development engagement.

The tools that do this well, n8n for orchestration, Evolution API for WhatsApp access, Chatwoot for unified inbox management, Supabase for structured data, and OpenAI for language intelligence, are either open-source or priced for small operators. The architecture to connect them is not complex. The value is in the design of the system, not the cost of the components. A clinic that understands this stops shopping for expensive software and starts building a system that actually works.

---

Frequently Asked Questions